Flexible Radial Analysis of Protoplanetary disks: FRAP

Flexible Radial Analysis of Protoplanetary disks (FRAP) version used for Yoshida et al. (2026). Summary Characterizing the physical properties of protoplanetary disks, such as surface density, temperature, and dust grain size, is critical for understanding planet formation. While multi-wavelength observations from interferometers like ALMA provide powerful insights, standard image-based analyses are frequently limited by biases due to a limited imaging beamsize. FRAP (Flexible Radial Analysis of Protoplanetary disks) provides an alternative, high-precision retrieval framework. Instead of relying on reconstructed images, FRAP directly models 1D visibilities. By representing the underlying physical structures as sample paths from Gaussian processes, it bypasses beam-size limitations and enables a flexible, unbiased extraction of radial disk profiles. Core Functions FRAP offers the following core capabilities: Direct Visibility Modeling: Compares 1D visibility models directly with observed interferometric data Non-Parametric Radial Profiling: Utilizes Gaussian processes to maintain high flexibility in modeling physical disk structures without strictly parameterized assumptions. Accelerated Bayesian Inference: Implemented entirely in JAX to allow for end-to-end automatic differentiation. Exact analytical gradients enable highly efficient posterior sampling via modern Markov-Chain Monte-Carlo (MCMC) technique, HMC/NUTS, using NumPyro. Dependencies FRAP is written in Python. Primary dependencies include: numpy scipy matploylib jax numpyro For comprehensive tutorials, API references, and usage examples, please visit our full documentation: https://frap.readthedocs.io/